Must-read on RL by Google DeepMind's Research Scientist Kevin Murphy dropped on ArXiv. It gives a clear, updated overview of deep RL and sequential decision-making, with examples.

Perhaps humans don't have the cognitive scaffolding to ever understand how these LLMs work.

Google just mapped every neuron & synapse in a block of mouse brain! 🤯 Slashing massively the cost and equipment barriers for such mapping. LICONN (Light Microscopy-based Connectomics) The electron microscopes used for connectomics research can cost millions of dollars, and

We're missing (at least one) major paradigm for LLM learning. Not sure what to call it, possibly it has a name - system prompt learning? Pretraining is for knowledge. Finetuning (SL/RL) is for habitual behavior. Both of these involve a change in parameters but a lot of human

LLMs haven't reached the level of autonomy so that they can be trusted with an entire profession, and it's already clear to everyone except for ignorant, deranged, or delusional people that they won't reach this level of autonomy. What remains is LLMs as a tool that humans use.

On the occasion of Feynman's birth anniversary, Here are all of his freely available lectures in one single thread 👇

The whole system prompt of Claude has been leaked on GitHub, 24,000 tokens long. It defines model behavior, tool use, and citation format.

The Claude 3.7 system prompt has been leaked and it's a goldmine for prompting techniques!

Intelligent Document Processing (IDP) Leaderboard – A Unified Benchmark for OCR, KIE, VQA, Table Extraction, and More. Its a unified benchmark for Vision-Language Models across 6 document AI tasks using 16 datasets and 9,229 documents. Gemini 2.5 Flash leads overall but stumbles

Neural Network:

Do you truly understand the limits of current AI?!

Stanford's Machine Learning Courses: ❯ CS221 - Artificial Intelligence ❯ CS229 - Machine Learning ❯ CS230 - Deep Learning ❯ CS234 - Reinforcement Learning ❯ CS224U - NL Understanding ❯ CS224N - NLP with Deep Learning All FREE courses. Links inside:

LLMs Get Lost in Multi-turn Conversation The cat is out of the bag. Pay attention, devs. This is one of the most common issues when building with LLMs today. Glad there is now paper to share insights. Here are my notes:

Why LLMs Fail in Back-and-Forth Chats Beautiful paper from @microsoft and Salesforce AI Research Large Language Models (LLMs) are incredibly good at tackling tasks when you give them all the information upfront in one go. Think of asking for a code snippet with all requirements clearly

That's a good entry point

Interested in the science of language models but tired of neural scaling laws? Here's a new perspective: our new paper presents neural thermodynamic laws -- thermodynamic concepts and laws naturally emerge in language model training! AI is naturAl, not Artificial, after all.

7. Introduction to Machine Learning Draw insights and offer recommendations from data, all while weighing the ethical dimensions of machine learning. Link: applieddigitalskills.withgoogle.com/c/middle-and-h…

mixture of experts (MoE) visually explained:

Wow. This is one of the best interactive sites I’ve seen for learning how LLMs work! 🔥 It starts w/ a clear intro and guides you through every core component: from Embedding, Layer Norm, and Self-Attention to MLPs, Transformer blocks, Softmax, and Output layers. link in 🧵↓

Flow matching produces smooth, deterministic trajectories. In contrast, the sampling process of a diffusion model is chaotic, resembling the random motion of gas particles.